Development and Validation of an ERCC1 Immunohistochemistry Assay for Solid Tumors

LUDLU-1 and NCI-H520 cells were selected as medium- and high-expressing cells, respectively, based on internal Western blot data (anti-ERCC1, AC-0206RUO; Epitomics, Burlingame, California). Cells were fixed in formalin and paraffin embedded. These cell-pellet blocks were used as run controls for the duration of the validation. Human tonsil and a breast cancer sample were used as tissue controls. One hundred twenty-five FFPE processed, surgically resected human tumor blocks were purchased commercially for use in this study (Asterand BioScience, Detroit, Michigan; ProteoGenex, Culver City, California). Tumor types included NSCLC (squamous cell and adenocarcinoma subtypes as 2 distinct categories), head and neck, esophageal, and ovarian cancer of varying histologic subtypes (Table 1). Additionally, FFPE blocks from 31 different healthy tissues were used to assess assay specificity.

Formalin-fixed, paraffin-embedded tissue samples were sectioned at 4 to 5 μm and placed onto positively charged glass slides. Sections were deparaffinized, rehydrated, and heat treated in a pressurized decloaking chamber (DC2008US; Biocare Medical, Concord, California) in EDTA buffer at 125°C, for 30 seconds and then cooled for 15 minutes to 95°C, for antigen retrieval. Slides were cooled to room temperature and incubated in a peroxidase blocking solution (Dako, Carpinteria, California) for 10 minutes. Slides were washed and incubated with the primary antibodies (ERCC1 and 4F9, OriGene, Rockville, Maryland; ERCC1 and D6G6, Cell Signaling Technology, Danvers, Massachusetts) for 1 hour at room temperature. Several iterations of antigen retrieval, primary antibody dilution, and incubation times were tested to determine the optimal conditions (Table 2). Simultaneously, a negative control-isotype slide was incubated with Universal Negative Mouse antibody (Dako) for 1 hour at room temperature. Detection was performed by incubation with polymer-horseradish peroxidase reagent (Dako) for 30 minutes, followed by incubated with 3,3′-diaminobenzidine chromogen for 5 minutes. Slides were counterstained with hematoxylin and coverslipped before being scanned using the iScan Coreo slide scanner (Ventana, Tucson, Arizona). All staining was performed manually.

The scoring system included a semiquantitative analysis of nuclear staining intensity. The staining intensity of ERCC1 was judged relative to the intensity of the internal stromal fibroblasts and/or endothelial cells as well as staining of the negative-control antibody. Internal positive controls immediately adjacent to the tumor area had to show positive staining for samples to be scored. Staining in the stromal cells was considered 2⁺, and the staining intensity of the tumor cells was scored in comparison to the level of 2⁺ exhibited in those adjacent stromal cells. The study pathologist (David W. Rogers, MD, LabCorp Clinical Trials) scored the samples by taking a minimum of 3 to 4 fields of view at the appropriate magnification (×40) and analyzed regions in which tumor and immediate adjacent stromal cells showed comparable reactivity. Outputs included both the percentage of positive cell results and an H score, a semiquantitative score derived by multiplying the percentage of the tumor at each intensity bin by the intensity number. This is illustrated in the formula below:

Similar to previous studies, the median H score per tumor type was used as a cutoff to define tumors ERCC1 positive or negative results.^4,6 A sample was considered ERCC1⁺ if the above internal control criteria were met and the assigned H score was greater than or equal to the median of the H scores for that specific tumor type. Negative staining of adjacent stromal cells resulted in a sample being scored not evaluable, regardless of tumor cell staining seen in the sample.

Initial screening was performed with 2 primary antibodies to ERCC1, clones 4F9 and D6G6, and a separate assay for each clone was optimized and developed (Table 2). Testing was done on FFPE blocks of 2 human tonsil cancers, 1 breast carcinoma, and a cell line with relatively high ERCC1 expression, NCI-520, based on Western blot analysis (data not shown). After optimization of each antibody, 25 FFPE NSCLC adenocarcinoma samples were stained with each antibody. Figure 1, A and B, shows ERCC1 staining with clone 4F9 and D6G6, respectively. After careful evaluation, it was deemed that the 4F9 clone provided a good signal to noise ratio and overall better ERCC1 staining in both cell-line controls and NSCLC adenocarcinoma samples (Figure 1). Even at high concentrations, the D6G6 clone staining was weak in all samples stained. From this evaluation, the 4F9 clone was moved forward for further evaluation and assay validation.

To better understand the ERCC1 expression across various solid tumors in which platinum therapy could be used for chemotherapy, we evaluated NSCLC, esophageal, head and neck, and ovarian tumor samples for ERCC1 expression using the 4F9 clone (Figure 2, A through E). Then, 125 surgically resected, FFPE-processed tumors were tested. The sample breakdown within each tumor type is listed in Table 3. From an initial cohort of 125 samples, 102 (82%) were deemed ERCC1 evaluable because of sufficient staining in the fibroblasts, and/or endothelial or stromal cells (Figure 3). The ERCC1⁺ and ERCC1⁻ samples for each tumor type is shown in Table 3. Overall, among the tumors tested, the frequency of ERCC1 positivity was similar to what has been previously reported (Table 4). There was no difference in staining patterns among the various tumor types.

Precision and reproducibility analyses were performed to determine the intra-assay and interassay variability, respectively, for the ERCC1 assay. Samples used here included the controls used in the optimization phase and a series of various tumor types that had been characterized for ERCC1 and that showed a range of expressions based on the H score (Figure 3). To assess assay precision, 5 serial sections from 12 tumor blocks and 5 controls were stained on one day. Similarly, to assess reproducibility, a slide from 11 tumor blocks and 5 controls was cut and stained on 5 different days. A mixed-effect model was used to obtain variance components for precision and reproducibility. In the mixed-effect model, variation in replicated runs from the same sample was treated as a random effect. The coefficient of variation (CV) for precision and reproducibility was derived from variance components for precision and reproducibility, respectively. Overall, we found the ERCC1 assay CV for precision (16.06%) and reproducibility (31.02%) to be acceptable (Table 3; Figures 4 and 5).

Specificity of the ERCC1 assay was assessed by testing the assay on a panel of normal human tissues. The assay exhibited minimal positive staining in this panel of tissues, with no unexpected findings. Of 93 samples tested, 3 (3.2%) showed positivity (H score ≥ 160, median of entire data set), and there was no obvious endothelium expression observed in healthy tissues (Table 5).

In this study, we performed a fit-for-purpose validation of an ERCC1 IHC assay that can be used to analyze samples in an exploratory setting. We found that the 4F9 clone provided the best ERCC1 signal, with no nonspecific staining among the 2 antibodies tested. We validated this ERCC1 assay using a variety of tumors types, including NSCLC, esophageal, head and neck, and ovarian cancer. The sensitivity analysis using this assay showed that there was a wide range of ERCC1 expression within the various tumor types (Table 4). This is consistent with previously published data that has shown that ERCC1-high and ERCC1-low populations existed within tumor types. Other groups have recently explored the use of this antibody clone to characterize the levels of ERCC1 in head and neck cancer as well as colorectal cancer.^4,18 

As part of the assay validation, we further evaluated parameters including precision, reproducibility, and specificity to ensure that the assay met the fit-for-purpose criteria. Published guidelines from the American Society of Clinical Oncology and the College of American Pathologists offer an in-depth literature review, coupled with recommendations to improve IHC assay specificity for established markers, such as HER2 across laboratories.^19,20  The College of American Pathologists Pathology and Laboratory Quality Center have provided recent information for the analytic validation of IHC assays meant to establish a novel assay or compare a novel assay to an existing assay.²¹  Bioanalytic assays, such as enzyme-linked immunosorbent assays, ligand-binding assays, or pharmacokinetic assays to measure drug exposure, have strict coefficient of variation acceptance levels in which assays should not exceed more than 20% for parameters such as precision and reproducibility during the validation.²²  These parameters are noted and recommended for testing in both the recent American Society of Clinical Oncology and College of American Pathologists publications, but strict acceptance criteria for CVs are not standard for IHC assays. Here, for the analytic testing, we used the same tumor sample set to measure precision and reproducibility. After thorough testing of both precision and reproducibility, we found that the CVs were low, 16.06% and 31.02%, respectively. In our experience, when validating IHC assays, higher CVs are observed in reproducibility testing than in precision; this can be attributed, in part, to reproducibility testing being performed to test the day-to-day variability, whereas precision testing is performed in one run.

In this study, we tested tumor samples that were procured from commercial sources. The advantage of using commercial sources is that tumor tissue can be collected at a reasonable cost and assay validations can be performed quickly. There are some inherent limitations with commercial tumor samples, including the inability to perform analyses, such as correlating ERCC1 protein levels and response because of a lack of available patient treatment and follow-up data. Additionally, these samples came from multiple clinical sites, and information about fixation time and processing methods was unavailable. However, the potential variety in tissue processing (e.g., fixation time) for these types of samples could mimic the variety of sample preparations seen when collecting samples from different clinical sites in a global study. In this study, the cell-line controls were generated under controlled fixation and processing conditions.

The nature of our scoring algorithm is similar to previous publications looking at ERCC1 expression, and potential stratification methods could include dividing groups into positive and negative results based on median or mean ERCC1 expression. The scoring here was based, in part, on the tumor tissue containing internal normal elements and the staining (intensity) of these structures when calculating the ERCC1 expression in the tumor area. A subset of tumor samples tested (n = 23, 18.4%) had no staining in fibroblasts or endothelial cells and were thus deemed not evaluable. Although the internal stromal controls may make a more stringent cutoff of the data, the staining and analysis of these normal elements allows for the control of differences in fixation.⁷  In potential clinical studies that would use this assay prospectively, because of the strict scoring criteria of internal-control staining, the screen failure rate would have to be monitored. In situations in which internal normal elements do not stain, testing fresh tumor biopsies could be an option to diminish the effects potentially caused by poor tissue processing.

This assay is now a validated, laboratory-developed test and can be used to measure ERCC1 protein levels in clinical trial–related FFPE samples. Transfer of this assay to another laboratory would require a bridge study that meets the concordance guidelines from the American Society of Clinical Oncology and the College of American Pathologists.¹⁹ 

We thank R. Smale, BS, and D. W. Rogers, MD, for their assistance in staining and analyzing the samples. We thank E. Lightcap, PhD, for providing Western blot data for ERCC1 on a panel of cell lines.